Date: Sun, 25 Feb 2007 11:31:36 +0100 From: Gabor Kovesdan <gabor@FreeBSD.org> To: Mark Linimon <linimon@lonesome.com> Cc: ports@freebsd.org Subject: Re: Optionally depending on one of two ports (or none of them) Message-ID: <45E16588.5040408@FreeBSD.org> In-Reply-To: <20070224201609.GC4960@soaustin.net> References: <20070222141301.007fee4f@localhost> <45DDA117.3050508@FreeBSD.org> <20070224163229.062bd234@localhost> <45E06810.1070704@FreeBSD.org> <45E08E3D.8060104@infracaninophile.co.uk> <20070224201609.GC4960@soaustin.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Mark Linimon schrieb: > On Sat, Feb 24, 2007 at 07:13:01PM +0000, Matthew Seaman wrote: > >> Why not do it this way? >> >> HAVE_TOR_DEVEL!= if pkg_info -I tor-devel-\* >/dev/null 2>&1 ; then echo YES; fi >> > > The problem with this approach is that you couple another shell invocation > into 'make index'. As long as they are in individual ports, it's ok, but > when we get them in the infrastructure, that's when things slow down. > > No, I don't have an alternative proposal. > > In theory, this is true, but some time ago when I worked on the Ports Collection infrestructure (during SoC), I did some tests about this. I removed some of the invocations from the common part of bsd.port.mk, which are not needed for the index building, but make does not have to know about this, thus such were evaluated as well during the index build. E.g. removing 3 entries means 3 times 15 000 - 16 000 shell invocations. Let's count with 15 000 ports, that means 45 000 shell invocations, and it made absolutely _no_ difference in the index building time even on a slower machine. Thus I think, this is not a problem in practice. Gabor
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?45E16588.5040408>